The nvimgcodec documentation recommends installing it using pip. I did not follow this recommendation and first because I was hoping to have the library as a “submodule” in my C++ project and just statically-link everything as much as I could. Eventually I found this wouldn’t be easy, and just proceeded to build and install it “by hand”. I was happy to do that, but then the question arrived of how to install it in the production board, where it would be better to have used the apt package (from nvImageCodec 0.7.0 Downloads | NVIDIA Developer, but I didn’t use it at first because of what I mentioned before). I just went ahead copying everything from `/opt`. Eventually I’ve met some challenges, described in this other thread (https://forums.developer.nvidia.com/t/nvimgcodec-produces-status-code-65535/).
Working on this issue I decided to go back and follow the recommendation of using the pip wheel. To me surprise, there was some change in the header file, the `nvimgcodecImageInfo_t` structure was missing the `buffer_size` attribute that the code example in the documentation uses.
Furthermore, if I go the pip way there’s no cmake files there, only a include folder.
What is the best way to install nvimgcodec for a C++ project?